Asymptotic Analysis and Singular Perturbation Theory
Asymptotic Analysis and Singular Perturbation Theory
Asymptotic Analysis and Singular Perturbation Theory
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
where<br />
w = − t2/3v .<br />
31/3 It follows that as t → ∞ with t2/3v fixed,<br />
u(x, t) ∼ 2π<br />
(3t) 1/3 <br />
f(0)Ai − t2/3v 31/3 <br />
.<br />
Thus the transition between oscillatory <strong>and</strong> exponential behavior is described by<br />
an Airy function. Since v = x/t, the width of the transition layer is of the order<br />
t 1/3 in x, <strong>and</strong> the solution in this region is of the order t −1/3 . Thus it decays more<br />
slowly <strong>and</strong> is larger than the solution elsewhere.<br />
Whitham [20] gives a detailed discussion of linear <strong>and</strong> nonlinear dispersive wave<br />
propagation.<br />
3.5 Laplace’s Method<br />
Consider an integral<br />
I(ε) =<br />
∞<br />
−∞<br />
f(t)e ϕ(t)/ε dt,<br />
where ϕ : R → R <strong>and</strong> f : R → C are smooth functions, <strong>and</strong> ε is a small positive<br />
parameter. This integral differs from the stationary phase integral in (3.9) because<br />
the argument of the exponential is real, not imaginary. Suppose that ϕ has a global<br />
maximum at t = c, <strong>and</strong> the maximum is nondegenerate, meaning that ϕ ′′ (c) < 0.<br />
The dominant contribution to the integral comes from the neighborhood of t = c,<br />
since the integr<strong>and</strong> is exponentially smaller in ε away from that point. Taylor<br />
exp<strong>and</strong>ing the functions in the integr<strong>and</strong> about t = c, we expect that<br />
<br />
I(ε) ∼<br />
Using the st<strong>and</strong>ard integral<br />
we get<br />
1<br />
[ϕ(c)+ f(c)e 2 ϕ′′ (c)(t−c) 2 ]/ε<br />
dt<br />
∼ f(c)e ϕ(c)/ε<br />
∞<br />
−∞<br />
∞<br />
−∞<br />
1 −<br />
e 2 at2<br />
<br />
2π<br />
dt =<br />
a ,<br />
<br />
2πε<br />
I(ε) ∼ f(c)<br />
|ϕ ′′ 1/2 e<br />
(c)|<br />
ϕ(c)/ε<br />
e 1<br />
2 ϕ′′ (c)(t−c) 2 /ε dt.<br />
as ε → 0 + .<br />
This result can proved under suitable assumptions on f <strong>and</strong> ϕ, but we will not give<br />
a detailed proof here (see [17], for example).<br />
43